| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Stephen Frost <sfrost(at)snowman(dot)net> |
| Cc: | Luke Lonergan <llonergan(at)greenplum(dot)com>, "Jim C(dot) Nasby" <jnasby(at)pervasive(dot)com>, Greg Stark <gsstark(at)mit(dot)edu>, Dann Corbit <DCorbit(at)connx(dot)com>, Simon Riggs <simon(at)2ndquadrant(dot)com>, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Merge algorithms for large numbers of "tapes" |
| Date: | 2006-03-09 23:59:42 |
| Message-ID: | 28664.1141948782@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Stephen Frost <sfrost(at)snowman(dot)net> writes:
> So, if we get a huge performance increase, what's wrong with:
> if [ sqrt(est(total)) <=3D work_mem ]; then
> two-pass-sort();
> else
> tape-sort();
> fi
> ?
Possibly nothing. However, from an algorithmic point of view the
CVS-tip code *is* two-pass-sort, given adequate work_mem and no
requirement for random access. Further, the available profile data
doesn't show any indication that the logtape.c code is eating 3/4ths
of the time (at least not after we fixed the ltsReleaseBlock problem).
So I basically do not believe Luke's assertion that removing logtape.c
is going to produce a 4X speedup. Maybe it's time to produce some code
that we can all test.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Luke Lonergan | 2006-03-10 00:04:48 | Re: Merge algorithms for large numbers of "tapes" |
| Previous Message | Dann Corbit | 2006-03-09 23:56:52 | Re: Merge algorithms for large numbers of "tapes" |